Title: Is Romance Dead? Generating Poetry using Character-Level Recurrent Neural Networks
Abstract: Recurrent Neural Networks (RNNs) are very useful for many natural language processing and language modeling tasks such as machine translation, parts-of-speech tagging, speech and handwriting recognition, semantic and sentiment analysis. The rapid development of machine learning led to a paradigm shift in how we model human languages. The application of RNNs to generate sequences of words to mimic that of actual human language is uncanny. In this talk, we present the power of RNNs to generate poetry - specifically haikus - by learning from a body of text taken from selected published research titles and abstracts of the UC Merced applied mathematics faculty and graduate students.